Negativity spreads faster: A large-scale multilingual twitter analysis on the role of sentiment in political communication

Antypas, Dimosthenis; Preece, Alun; Camacho-Collados, José

doi:10.1016/j.osnem.2023.100242

Cited by 24 publications

(16 citation statements)

References 37 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Additionally, this study also investigates the methodological utility of comments from political vloggers' channels. Nowadays, NLP via machine learning is an emerging method for political communication research, and a compelling area in NLP research is text classification, which enables us to distinguish the political leaning of text [e.g., [28][29][30]. In the process of classifier development, necessary and important resources are training data for machine learning; the quality of training data can heavily influence the overall accuracy of analysis.…”

Section: The Current Studymentioning

confidence: 99%

Where do cross-cutting discussions happen?: Identifying cross-cutting comments on YouTube videos of political vloggers and mainstream news outlets

Chae,

Lee

2024

PLoS ONE

View full text Add to dashboard Cite

Since the conception of social media, research on political communication has pointed toward the risk that the social media environment can foster political echo chambers. However, this has recently been contradicted by some studies demonstrating “cross-cutting discussions” on social media. The current study extends this literature by particularly focusing on communication on political vlogger videos and having mainstream news outlet videos as a reference point. Specifically, this study addresses five points: (1) to what extent cross-partisan comments occupy conservative and liberal vloggers’ comment threads and if there is a significant difference between the two, (2) the possibility that comments from vlogger videos can be utilized to predict the political leanings of comments on mainstream news outlet videos, (3) if the proportion of cross-cutting discussions on mainstream news outlet videos significantly varies by the news outlet’s political leaning, (4) if a neutral news outlet channel can work as a venue for cross-cutting discussions, and (5) if the proportion of cross-cutting comments in mainstream news outlet comment threads is significantly different from that in vlogger comment threads. Both manual and computational analyses were employed; the political leanings of vlogger comments were analyzed by manual content analysis, and based on the results, the political leanings of mainstream news outlet comments were analyzed by NLP classifiers using three different algorithms—logistic regression, SVM, and random forest. As a result, we found that the proportion of cross-cutting discussions significantly varies by both the channel’s political leaning and media type. In addition, our results suggest the possibility of neutral news outlets as a place for cross-cutting discussions.

show abstract

Section: The Current Studymentioning

confidence: 99%

Where do cross-cutting discussions happen?: Identifying cross-cutting comments on YouTube videos of political vloggers and mainstream news outlets

Chae,

Lee

2024

PLoS ONE

View full text Add to dashboard Cite

show abstract

“…Consequently, political campaigns have become increasingly negative over time (Haselmayer et al, 2019;Külz et al, 2023). This trend has been partially attributed to negativity bias, as individuals tend to pay more attention to negative content than positive information (Antypas et al, 2023;Meffert et al, 2006;Schöne et al, 2021). Furthermore, negative content is more likely to attract the attention of news media, particularly in the political context (Haselmayer et al, 2019).…”

Section: In the Realms Of Negative Campaigningmentioning

confidence: 99%

“…The transition towards a more aggressive political discourse is particularly noticeable in an era characterized by the dominance of social media. Previous studies show that spaces where users can have prompt and direct content dissemination and communication are more likely to be associated with negative content like Twitter (Antypas et al, 2023;Schöne et al, 2021). This trend can potentially diminish the quality of democratic discussions by marginalizing substantive debates and citizens' political participation (Klinger et al, 2023;Mubarok, 2022).…”

Section: In the Realms Of Negative Campaigningmentioning

confidence: 99%

Evolution of Brazilian Democracy: Unveiling Election Dynamics in Political Issues, Negativity, and Acclaim

Gonçalves,

De-Lima-Santos,

Fenoll

et al. 2024

PaG

View full text Add to dashboard Cite

In recent years, Brazil, as the world’s fourth-largest democracy, witnessed the dominance of polarized and symbolically charged electoral campaigns on social media, culminating in the election of a populist political figure in 2018 and his subsequent defeat in 2022. Extensive research has indicated that political campaigns often sidelined substantive policy proposals in favor of negative and divisive issues. However, a critical gap remains in the absence of temporal investigations contrasting the prevalence of negativity and acclaim campaigns on social media platforms during elections. This study addresses this gap by examining associations between political issues and negative and acclaim campaigns across two Brazilian electoral campaigns. Drawing upon a sample of messages posted on Twitter (n = 1,191) during the presidential elections of 2018 and 2022, our study reveals associations between substantive political issues, such as education and health, and acclaim campaign strategies, while the divisive issues of Covid-19 and corruption are associated with negative campaign strategies. Moreover, the results suggest that gender policy is related to both acclaim and negative messages since it is a polarizing issue in Brazilian politics. Our study also shows an increased negativity trend, with the 2022 presidential election campaign more likely to be negative than in 2018. By conducting a temporal analysis of Brazil’s political context, our study sheds light on the evolving dynamics of political communication in the age of social media, contributing substantially to the literature on negativity in political campaigns.

show abstract

“…Experiments by [21]- [25] mention that the current topic-guided language models generate problems that are conceivably coherent compared to those of the regular Latent Dirichlet Allocation (LDA) topic model. The authors of the studies [26] - [29] propose a text feature extraction algorithm based on a deep neural network that improves the efficiency of modern social media. Generative Adversarial Networks (GAN) have received much attention and produced impressive results.…”

Section: Related Workmentioning

confidence: 99%

“…April 2023 Revised:12 June 2023 Accepted:29 June 2023 ___________________________________________________________________________________________________________________ categories than the Naïve Bayes Classifier. The accuracy scores for all crime categories are relatively close for the SVM Classifier and Random Forest Classifier, indicating their effectiveness in classifying crime-related tweets.…”

mentioning

confidence: 99%

Expressivity of Tweets on Social Issues Using Aspect Based Text Classification

Thomas,

Parthasarathy

2023

IJRITCC

View full text Add to dashboard Cite

Social discussions about crime on Twitter and open forums aims to understand the barriers that hinder people from expressing their concerns or aligning with popular opinions. A curated dataset spanning three months in 2023 was collected, covering categories like crimes and Gender Equality and Violence Against Women. The study employs aspect-based sentiment analysis to classify sentiment polarity in tweets, utilising a comprehensive framework involving three text feature classification stages. The initial stage analyses individual words, phrases, and tweet patterns to classify text features based on specific linguistic elements. In the subsequent step, semantic relations explore a better understanding of the core sentiment and infer relationships between different text keywords. This stage enhances the analysis by considering the meaning and contextual nuances of the language used in the tweets. The final stage incorporates transformer-based models for effective multilabel classification to view the diversity present in the dataset. The study's quantitative analysis reveals that the Ensemble learning approach demonstrates an impressive precision measure of 93%. By integrating the three stages of text feature classification, the study enhances the accuracy and comprehensiveness of sentiment analysis in social discussions about crime on Twitter.

show abstract

Negativity spreads faster: A large-scale multilingual twitter analysis on the role of sentiment in political communication

Cited by 24 publications

References 37 publications

Where do cross-cutting discussions happen?: Identifying cross-cutting comments on YouTube videos of political vloggers and mainstream news outlets

Where do cross-cutting discussions happen?: Identifying cross-cutting comments on YouTube videos of political vloggers and mainstream news outlets

Evolution of Brazilian Democracy: Unveiling Election Dynamics in Political Issues, Negativity, and Acclaim

Expressivity of Tweets on Social Issues Using Aspect Based Text Classification

Contact Info

Product

Resources

About